Two-stage network meta-regression for heterogeneous treatment effects
This method helps estimate how well different treatments work for individual patients by combining information about their personal risk of getting worse with results from multiple treatment studies. It uses two steps: first, predicting a patient's baseline risk of a bad outcome using their personal traits; second, using that risk score to see how treatment benefits change depending on how high or low the risk is.
At a glance
Use when
Comparing multiple treatments in a network meta-analysis while accounting for patient-level risk heterogeneity; aiming to support individualized clinical decisions using prognostic information.
Avoid when
Only aggregate study-level data are available; when individual patient data are too sparse or incomplete; when no clear prognostic factors exist for the outcome.
Inputs
Individual patient data including baseline characteristics and outcomes from randomized clinical trials; treatment assignment; outcome of interest; network of interventions being compared.
Outputs
Baseline risk scores for patients; risk-stratified relative and absolute treatment effects; personalized treatment effect estimates across a network of interventions.
How it works
A two-stage statistical method combining prognostic modeling and network meta-regression to estimate heterogeneous treatment effects. In stage one, a prognostic model is developed using individual patient data to predict baseline risk of the outcome. In stage two, the derived baseline risk score is used as an effect modifier in a network meta-regression model to assess how relative and absolute treatment effects vary across risk levels. The method enables personalized treatment recommendations within a network meta-analysis framework.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Technology
- Non-specific
- Assumptions
- The baseline risk score adequately captures patient prognosis; the relationship between baseline risk and treatment effect is consistent across trials; linearity or specified functional form in the meta-regression; availability of individual patient data for model development.
- Strengths
- Enables personalized treatment recommendations; integrates prognosis and treatment evidence; uses individual patient data to improve precision; allows exploration of treatment effect heterogeneity beyond single covariates.
- Limitations
- Requires individual patient data which may not be available; results depend on quality of the prognostic model; assumes baseline risk is a valid effect modifier; limited generalizability if patient populations are narrow.
- Also known as
- Two-stage prediction model for heterogeneous treatment effects, Baseline risk-stratified network meta-regression
Questions this answers
- › How do treatment effects vary for patients with different baseline risks?
- › Which treatment works best for high-risk versus low-risk patients?
- › How do patient characteristics like age or disability affect treatment benefit?
- › Can we predict the best treatment for an individual patient based on their risk profile?
- › Does the effectiveness of multiple sclerosis treatments depend on a patient's risk of relapse?
- › How can we combine prognosis models with treatment comparison studies?
Related methods
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.